Cataract Detection by Leveraging VGG-19 Classification Model on Retinal Images

A cataract is an obscurity of the eye lens as a result of protein buildup in the lens. Almost 48% of the world’s blindness and 51% of South Asia’s blindness are a result of cataract formation. Hence, it has become necessary to tackle this issue quickly. An individual’s rate of blindness and surgery pain could be reduced upon timely detection. This paper proposes a modified version of the VGG-19 model to detect cataracts in retinal fundus images. Additional layers are added to the original VGG-19 model to obtain the proposed model. A preprocessed dataset is split into training, testing and validation sections. The proposed model is then trained over the training data, which is further validated and an accuracy of 99.89% is observed. Finally, the testing data is fed to the model and a high testing accuracy of 97.25% is obtained.